: p-value in statistics
In: 10
For folks who still find these explanations overwhelming, a very small p-value is an indication that the data is true. The smaller the number and closer to zero, the less doubt there is.
So if you see a statement like “four out of five people with blonde hair are actually brunettes who bleach their hair (p<0.1)” that is not a *great* indication that most blondes are faking it. If that number changes to (p<0.0005), that’s actually a good sign that it’s true (I made those stats up, not real, obvs).
The important thing to know is that the p-value is derived from the existing *data*, and how the pure numbers relate to each other, not reality. If you knew that the above second statistic was based on information gathered at a hair salon that specializes in hair coloring, then maybe you can’t apply that factoid to the real world, just people who go to a salon.
So, a p-value is basically quality control of a data set and its result. But you need to know how the data was obtained and under what conditions before you should just accept a statement because of an accompanying small p-value. In science we are trying to move away from p-values for many reasons, but partly because you can hack the data to get a number you want and therefore claim whatever you want. “The p-value is good so it must be true!”
Latest Answers